Car Detection in aerial imagery¶
After a massive research I chose to work with Detectron2 object detection model of Meta.
I took the following steps:
- Install the model dependencies
- Download my custom dataset
- Write our model training configuration
- Run the model training
- Evaluate the model performance
- Run the model inference on test images
- Convert image to green
- Paint bboxes red
- Execute the desired algorithm
Install Dependencies¶
In [ ]:
# install dependencies: (use cu101 because colab has CUDA 10.1)
!pip install -U torch==1.5 torchvision==0.6 -f https://download.pytorch.org/whl/cu101/torch_stable.html
!pip install cython pyyaml==5.1
!pip install -U 'git+https://github.com/cocodataset/cocoapi.git#subdirectory=PythonAPI'
import torch, torchvision
print(torch.__version__, torch.cuda.is_available())
!gcc --version
# install detectron2:
!pip install detectron2==0.1.3 -f https://dl.fbaipublicfiles.com/detectron2/wheels/cu101/torch1.5/index.html
Looking in indexes: https://pypi.org/simple, https://us-python.pkg.dev/colab-wheels/public/simple/
Looking in links: https://download.pytorch.org/whl/cu101/torch_stable.html
Collecting torch==1.5
Downloading https://download.pytorch.org/whl/cu101/torch-1.5.0%2Bcu101-cp37-cp37m-linux_x86_64.whl (703.8 MB)
|████████████████████████████████| 703.8 MB 20 kB/s
Collecting torchvision==0.6
Downloading https://download.pytorch.org/whl/cu101/torchvision-0.6.0%2Bcu101-cp37-cp37m-linux_x86_64.whl (6.6 MB)
|████████████████████████████████| 6.6 MB 1.3 MB/s
Requirement already satisfied: future in /usr/local/lib/python3.7/dist-packages (from torch==1.5) (0.16.0)
Requirement already satisfied: numpy in /usr/local/lib/python3.7/dist-packages (from torch==1.5) (1.21.6)
Requirement already satisfied: pillow>=4.1.1 in /usr/local/lib/python3.7/dist-packages (from torchvision==0.6) (7.1.2)
Installing collected packages: torch, torchvision
Attempting uninstall: torch
Found existing installation: torch 1.12.1+cu113
Uninstalling torch-1.12.1+cu113:
Successfully uninstalled torch-1.12.1+cu113
Attempting uninstall: torchvision
Found existing installation: torchvision 0.13.1+cu113
Uninstalling torchvision-0.13.1+cu113:
Successfully uninstalled torchvision-0.13.1+cu113
ERROR: pip's dependency resolver does not currently take into account all the packages that are installed. This behaviour is the source of the following dependency conflicts.
torchtext 0.13.1 requires torch==1.12.1, but you have torch 1.5.0+cu101 which is incompatible.
torchaudio 0.12.1+cu113 requires torch==1.12.1, but you have torch 1.5.0+cu101 which is incompatible.
fastai 2.7.9 requires torch<1.14,>=1.7, but you have torch 1.5.0+cu101 which is incompatible.
fastai 2.7.9 requires torchvision>=0.8.2, but you have torchvision 0.6.0+cu101 which is incompatible.
Successfully installed torch-1.5.0+cu101 torchvision-0.6.0+cu101
Looking in indexes: https://pypi.org/simple, https://us-python.pkg.dev/colab-wheels/public/simple/
Requirement already satisfied: cython in /usr/local/lib/python3.7/dist-packages (0.29.32)
Collecting pyyaml==5.1
Downloading PyYAML-5.1.tar.gz (274 kB)
|████████████████████████████████| 274 kB 2.1 MB/s
Building wheels for collected packages: pyyaml
Building wheel for pyyaml (setup.py) ... done
Created wheel for pyyaml: filename=PyYAML-5.1-cp37-cp37m-linux_x86_64.whl size=44092 sha256=629d2ac28b2aabdf6da382a0e6455ae18763d5114eb8c6d16dc87a7ceb398ca9
Stored in directory: /root/.cache/pip/wheels/77/f5/10/d00a2bd30928b972790053b5de0c703ca87324f3fead0f2fd9
Successfully built pyyaml
Installing collected packages: pyyaml
Attempting uninstall: pyyaml
Found existing installation: PyYAML 6.0
Uninstalling PyYAML-6.0:
Successfully uninstalled PyYAML-6.0
ERROR: pip's dependency resolver does not currently take into account all the packages that are installed. This behaviour is the source of the following dependency conflicts.
fastai 2.7.9 requires torch<1.14,>=1.7, but you have torch 1.5.0+cu101 which is incompatible.
fastai 2.7.9 requires torchvision>=0.8.2, but you have torchvision 0.6.0+cu101 which is incompatible.
dask 2022.2.0 requires pyyaml>=5.3.1, but you have pyyaml 5.1 which is incompatible.
Successfully installed pyyaml-5.1
Looking in indexes: https://pypi.org/simple, https://us-python.pkg.dev/colab-wheels/public/simple/
Collecting git+https://github.com/cocodataset/cocoapi.git#subdirectory=PythonAPI
Cloning https://github.com/cocodataset/cocoapi.git to /tmp/pip-req-build-fwarmm91
Running command git clone -q https://github.com/cocodataset/cocoapi.git /tmp/pip-req-build-fwarmm91
Requirement already satisfied: setuptools>=18.0 in /usr/local/lib/python3.7/dist-packages (from pycocotools==2.0) (57.4.0)
Requirement already satisfied: cython>=0.27.3 in /usr/local/lib/python3.7/dist-packages (from pycocotools==2.0) (0.29.32)
Requirement already satisfied: matplotlib>=2.1.0 in /usr/local/lib/python3.7/dist-packages (from pycocotools==2.0) (3.2.2)
Requirement already satisfied: cycler>=0.10 in /usr/local/lib/python3.7/dist-packages (from matplotlib>=2.1.0->pycocotools==2.0) (0.11.0)
Requirement already satisfied: numpy>=1.11 in /usr/local/lib/python3.7/dist-packages (from matplotlib>=2.1.0->pycocotools==2.0) (1.21.6)
Requirement already satisfied: kiwisolver>=1.0.1 in /usr/local/lib/python3.7/dist-packages (from matplotlib>=2.1.0->pycocotools==2.0) (1.4.4)
Requirement already satisfied: pyparsing!=2.0.4,!=2.1.2,!=2.1.6,>=2.0.1 in /usr/local/lib/python3.7/dist-packages (from matplotlib>=2.1.0->pycocotools==2.0) (3.0.9)
Requirement already satisfied: python-dateutil>=2.1 in /usr/local/lib/python3.7/dist-packages (from matplotlib>=2.1.0->pycocotools==2.0) (2.8.2)
Requirement already satisfied: typing-extensions in /usr/local/lib/python3.7/dist-packages (from kiwisolver>=1.0.1->matplotlib>=2.1.0->pycocotools==2.0) (4.1.1)
Requirement already satisfied: six>=1.5 in /usr/local/lib/python3.7/dist-packages (from python-dateutil>=2.1->matplotlib>=2.1.0->pycocotools==2.0) (1.15.0)
Building wheels for collected packages: pycocotools
Building wheel for pycocotools (setup.py) ... done
Created wheel for pycocotools: filename=pycocotools-2.0-cp37-cp37m-linux_x86_64.whl size=265163 sha256=760ad144520e45f46cece7934bb8a986cb018813ce830d8bb9aa992b86a21558
Stored in directory: /tmp/pip-ephem-wheel-cache-lulpwcul/wheels/e2/6b/1d/344ac773c7495ea0b85eb228bc66daec7400a143a92d36b7b1
Successfully built pycocotools
Installing collected packages: pycocotools
Attempting uninstall: pycocotools
Found existing installation: pycocotools 2.0.4
Uninstalling pycocotools-2.0.4:
Successfully uninstalled pycocotools-2.0.4
Successfully installed pycocotools-2.0
1.5.0+cu101 True
gcc (Ubuntu 7.5.0-3ubuntu1~18.04) 7.5.0
Copyright (C) 2017 Free Software Foundation, Inc.
This is free software; see the source for copying conditions. There is NO
warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
Looking in indexes: https://pypi.org/simple, https://us-python.pkg.dev/colab-wheels/public/simple/
Looking in links: https://dl.fbaipublicfiles.com/detectron2/wheels/cu101/torch1.5/index.html
Collecting detectron2==0.1.3
Downloading https://dl.fbaipublicfiles.com/detectron2/wheels/cu101/torch1.5/detectron2-0.1.3%2Bcu101-cp37-cp37m-linux_x86_64.whl (6.2 MB)
|████████████████████████████████| 6.2 MB 1.5 MB/s
Requirement already satisfied: future in /usr/local/lib/python3.7/dist-packages (from detectron2==0.1.3) (0.16.0)
Requirement already satisfied: tqdm>4.29.0 in /usr/local/lib/python3.7/dist-packages (from detectron2==0.1.3) (4.64.1)
Requirement already satisfied: Pillow in /usr/local/lib/python3.7/dist-packages (from detectron2==0.1.3) (7.1.2)
Requirement already satisfied: tensorboard in /usr/local/lib/python3.7/dist-packages (from detectron2==0.1.3) (2.8.0)
Requirement already satisfied: pydot in /usr/local/lib/python3.7/dist-packages (from detectron2==0.1.3) (1.3.0)
Requirement already satisfied: tabulate in /usr/local/lib/python3.7/dist-packages (from detectron2==0.1.3) (0.8.10)
Requirement already satisfied: termcolor>=1.1 in /usr/local/lib/python3.7/dist-packages (from detectron2==0.1.3) (1.1.0)
Collecting yacs>=0.1.6
Downloading yacs-0.1.8-py3-none-any.whl (14 kB)
Requirement already satisfied: matplotlib in /usr/local/lib/python3.7/dist-packages (from detectron2==0.1.3) (3.2.2)
Collecting fvcore>=0.1.1
Downloading fvcore-0.1.5.post20220512.tar.gz (50 kB)
|████████████████████████████████| 50 kB 1.6 MB/s
Collecting mock
Downloading mock-4.0.3-py3-none-any.whl (28 kB)
Requirement already satisfied: cloudpickle in /usr/local/lib/python3.7/dist-packages (from detectron2==0.1.3) (1.5.0)
Requirement already satisfied: numpy in /usr/local/lib/python3.7/dist-packages (from fvcore>=0.1.1->detectron2==0.1.3) (1.21.6)
Requirement already satisfied: pyyaml>=5.1 in /usr/local/lib/python3.7/dist-packages (from fvcore>=0.1.1->detectron2==0.1.3) (5.1)
Collecting iopath>=0.1.7
Downloading iopath-0.1.10.tar.gz (42 kB)
|████████████████████████████████| 42 kB 1.0 MB/s
Requirement already satisfied: typing_extensions in /usr/local/lib/python3.7/dist-packages (from iopath>=0.1.7->fvcore>=0.1.1->detectron2==0.1.3) (4.1.1)
Collecting portalocker
Downloading portalocker-2.5.1-py2.py3-none-any.whl (15 kB)
Requirement already satisfied: kiwisolver>=1.0.1 in /usr/local/lib/python3.7/dist-packages (from matplotlib->detectron2==0.1.3) (1.4.4)
Requirement already satisfied: cycler>=0.10 in /usr/local/lib/python3.7/dist-packages (from matplotlib->detectron2==0.1.3) (0.11.0)
Requirement already satisfied: pyparsing!=2.0.4,!=2.1.2,!=2.1.6,>=2.0.1 in /usr/local/lib/python3.7/dist-packages (from matplotlib->detectron2==0.1.3) (3.0.9)
Requirement already satisfied: python-dateutil>=2.1 in /usr/local/lib/python3.7/dist-packages (from matplotlib->detectron2==0.1.3) (2.8.2)
Requirement already satisfied: six>=1.5 in /usr/local/lib/python3.7/dist-packages (from python-dateutil>=2.1->matplotlib->detectron2==0.1.3) (1.15.0)
Requirement already satisfied: protobuf>=3.6.0 in /usr/local/lib/python3.7/dist-packages (from tensorboard->detectron2==0.1.3) (3.17.3)
Requirement already satisfied: werkzeug>=0.11.15 in /usr/local/lib/python3.7/dist-packages (from tensorboard->detectron2==0.1.3) (1.0.1)
Requirement already satisfied: grpcio>=1.24.3 in /usr/local/lib/python3.7/dist-packages (from tensorboard->detectron2==0.1.3) (1.48.1)
Requirement already satisfied: requests<3,>=2.21.0 in /usr/local/lib/python3.7/dist-packages (from tensorboard->detectron2==0.1.3) (2.23.0)
Requirement already satisfied: tensorboard-plugin-wit>=1.6.0 in /usr/local/lib/python3.7/dist-packages (from tensorboard->detectron2==0.1.3) (1.8.1)
Requirement already satisfied: tensorboard-data-server<0.7.0,>=0.6.0 in /usr/local/lib/python3.7/dist-packages (from tensorboard->detectron2==0.1.3) (0.6.1)
Requirement already satisfied: absl-py>=0.4 in /usr/local/lib/python3.7/dist-packages (from tensorboard->detectron2==0.1.3) (1.2.0)
Requirement already satisfied: google-auth-oauthlib<0.5,>=0.4.1 in /usr/local/lib/python3.7/dist-packages (from tensorboard->detectron2==0.1.3) (0.4.6)
Requirement already satisfied: google-auth<3,>=1.6.3 in /usr/local/lib/python3.7/dist-packages (from tensorboard->detectron2==0.1.3) (1.35.0)
Requirement already satisfied: wheel>=0.26 in /usr/local/lib/python3.7/dist-packages (from tensorboard->detectron2==0.1.3) (0.37.1)
Requirement already satisfied: markdown>=2.6.8 in /usr/local/lib/python3.7/dist-packages (from tensorboard->detectron2==0.1.3) (3.4.1)
Requirement already satisfied: setuptools>=41.0.0 in /usr/local/lib/python3.7/dist-packages (from tensorboard->detectron2==0.1.3) (57.4.0)
Requirement already satisfied: cachetools<5.0,>=2.0.0 in /usr/local/lib/python3.7/dist-packages (from google-auth<3,>=1.6.3->tensorboard->detectron2==0.1.3) (4.2.4)
Requirement already satisfied: pyasn1-modules>=0.2.1 in /usr/local/lib/python3.7/dist-packages (from google-auth<3,>=1.6.3->tensorboard->detectron2==0.1.3) (0.2.8)
Requirement already satisfied: rsa<5,>=3.1.4 in /usr/local/lib/python3.7/dist-packages (from google-auth<3,>=1.6.3->tensorboard->detectron2==0.1.3) (4.9)
Requirement already satisfied: requests-oauthlib>=0.7.0 in /usr/local/lib/python3.7/dist-packages (from google-auth-oauthlib<0.5,>=0.4.1->tensorboard->detectron2==0.1.3) (1.3.1)
Requirement already satisfied: importlib-metadata>=4.4 in /usr/local/lib/python3.7/dist-packages (from markdown>=2.6.8->tensorboard->detectron2==0.1.3) (4.12.0)
Requirement already satisfied: zipp>=0.5 in /usr/local/lib/python3.7/dist-packages (from importlib-metadata>=4.4->markdown>=2.6.8->tensorboard->detectron2==0.1.3) (3.8.1)
Requirement already satisfied: pyasn1<0.5.0,>=0.4.6 in /usr/local/lib/python3.7/dist-packages (from pyasn1-modules>=0.2.1->google-auth<3,>=1.6.3->tensorboard->detectron2==0.1.3) (0.4.8)
Requirement already satisfied: idna<3,>=2.5 in /usr/local/lib/python3.7/dist-packages (from requests<3,>=2.21.0->tensorboard->detectron2==0.1.3) (2.10)
Requirement already satisfied: chardet<4,>=3.0.2 in /usr/local/lib/python3.7/dist-packages (from requests<3,>=2.21.0->tensorboard->detectron2==0.1.3) (3.0.4)
Requirement already satisfied: certifi>=2017.4.17 in /usr/local/lib/python3.7/dist-packages (from requests<3,>=2.21.0->tensorboard->detectron2==0.1.3) (2022.6.15)
Requirement already satisfied: urllib3!=1.25.0,!=1.25.1,<1.26,>=1.21.1 in /usr/local/lib/python3.7/dist-packages (from requests<3,>=2.21.0->tensorboard->detectron2==0.1.3) (1.24.3)
Requirement already satisfied: oauthlib>=3.0.0 in /usr/local/lib/python3.7/dist-packages (from requests-oauthlib>=0.7.0->google-auth-oauthlib<0.5,>=0.4.1->tensorboard->detectron2==0.1.3) (3.2.0)
Building wheels for collected packages: fvcore, iopath
Building wheel for fvcore (setup.py) ... done
Created wheel for fvcore: filename=fvcore-0.1.5.post20220512-py3-none-any.whl size=61288 sha256=b40b4e9c0739515556fb31939b628f8f8c3b6d0a803164e6fe99045edfbb3f35
Stored in directory: /root/.cache/pip/wheels/68/20/f9/a11a0dd63f4c13678b2a5ec488e48078756505c7777b75b29e
Building wheel for iopath (setup.py) ... done
Created wheel for iopath: filename=iopath-0.1.10-py3-none-any.whl size=31549 sha256=134bb754194a23c923c5a89dfe19e8230d09ca07b77ff4fcae5f7cb38b66e2f1
Stored in directory: /root/.cache/pip/wheels/aa/cc/ed/ca4e88beef656b01c84b9185196513ef2faf74a5a379b043a7
Successfully built fvcore iopath
Installing collected packages: portalocker, yacs, iopath, mock, fvcore, detectron2
Successfully installed detectron2-0.1.3+cu101 fvcore-0.1.5.post20220512 iopath-0.1.10 mock-4.0.3 portalocker-2.5.1 yacs-0.1.8
Relevant imports¶
In [ ]:
# You may need to restart your runtime prior to this, to let your installation take effect
# Some basic setup:
# Setup detectron2 logger
import detectron2
from detectron2.utils.logger import setup_logger
setup_logger()
# import some common libraries
import numpy as np
import cv2
import random
from google.colab.patches import cv2_imshow
import random
import os
import glob
# import some common detectron2 utilities
from detectron2 import model_zoo
from detectron2.engine import DefaultPredictor, DefaultTrainer
from detectron2.config import get_cfg
from detectron2.utils.visualizer import Visualizer, ColorMode
from detectron2.data import DatasetCatalog, MetadataCatalog, build_detection_test_loader
from detectron2.data.catalog import DatasetCatalog
from detectron2.data.datasets import register_coco_instances
from detectron2.evaluation import COCOEvaluator, inference_on_dataset
from detectron2.modeling import build_model
Import and Register my custom Dataset¶
Note: if the link will expire let me know and I'll send a new one
In [ ]:
!curl -L "https://app.roboflow.com/ds/CGdYIqBsHg?key=2WZlODtuOg" > roboflow.zip; unzip roboflow.zip; rm roboflow.zip
!rm -f *.txt
% Total % Received % Xferd Average Speed Time Time Time Current
Dload Upload Total Spent Left Speed
100 892 100 892 0 0 1429 0 --:--:-- --:--:-- --:--:-- 1429
100 5723k 100 5723k 0 0 3493k 0 0:00:01 0:00:01 --:--:-- 7083k
Archive: roboflow.zip
extracting: README.dataset.txt
extracting: README.roboflow.txt
creating: test/
extracting: test/00000266_co_png.rf.d52cbe2b4b992346d4009cff1f923f61.jpg
extracting: test/00000268_co_png.rf.adada98eba89c01aa93258fcfdc03937.jpg
extracting: test/00000271_co_png.rf.07fc054292145538bb081e9d8e718cdb.jpg
extracting: test/00000272_co_png.rf.c52618161985e36efb1e5528e3b379b5.jpg
extracting: test/00000274_co_png.rf.cd935e4a3b74749b7f37e3c66a41a9ea.jpg
extracting: test/00000275_co_png.rf.cc1d34e2a615d0d4496e3486e1acdef3.jpg
extracting: test/00000276_co_png.rf.a196ad96b2a8a80273dd3d58f7a3c738.jpg
extracting: test/00000278_co_png.rf.000b56364382882add11e6eaf41a5e5b.jpg
extracting: test/00000279_co_png.rf.6928e406455ac417a308d36a1c2d16ba.jpg
extracting: test/00000280_co_png.rf.d5376a828d7bb60dd639da740407393b.jpg
extracting: test/00000281_co_png.rf.1926133066b5e717081489f820104f0c.jpg
extracting: test/00000282_co_png.rf.03941b08108d78b8299a94de94ef19c2.jpg
extracting: test/00000283_co_png.rf.f702d5b45319ca59c514421cd511e5ad.jpg
extracting: test/00000284_co_png.rf.3617d46739297d025037e30fc93f75b1.jpg
extracting: test/00000285_co_png.rf.31310bf523d2e2b264535f06028b2183.jpg
extracting: test/00000286_co_png.rf.63734907043a5028cba27d42902c98f0.jpg
extracting: test/00000287_co_png.rf.799e739c7f68e74bd2ed61da0ded4027.jpg
extracting: test/00000288_co_png.rf.0173c05a5c92a1a38ce0e8297b62656e.jpg
extracting: test/00000289_co_png.rf.3a4d97d6d595c59b84a4350557025436.jpg
extracting: test/00000293_co_png.rf.873a2fd2cceabe448f7d63218f773e1c.jpg
extracting: test/00000295_co_png.rf.ecfddb62d87bf3d1093f427bb6528f89.jpg
extracting: test/00000298_co_png.rf.01da5121d96add751e9576b58ba26777.jpg
extracting: test/00000299_co_png.rf.5001a083eb60c1b3b0c4799691d9e117.jpg
extracting: test/00000300_co_png.rf.17ff89cf9978650a6085c8e74b5d7448.jpg
extracting: test/00000302_co_png.rf.0fc96a833631ec1bb79e8b0814b372e1.jpg
extracting: test/00000303_co_png.rf.5e54caeba2cfea103a37004c0e4ba0cb.jpg
extracting: test/00000307_co_png.rf.aefdc6cacd2413471a197b1d889e0685.jpg
extracting: test/00000308_co_png.rf.f3e056da60b4e9d8660ae905f87ff42b.jpg
extracting: test/00000309_co_png.rf.17e79d1ac5ad002588512b3de0ed9eb6.jpg
extracting: test/00000312_co_png.rf.a73985f20b42c36570acbd95b4c8600c.jpg
extracting: test/00000315_co_png.rf.6de4a46482e928403e6c5c329356d4d7.jpg
extracting: test/00000317_co_png.rf.2e8da912578d384d6c87e08aeeb4addc.jpg
extracting: test/00000318_co_png.rf.29577a521b20a72e5947d3b4b3e4766b.jpg
extracting: test/00000319_co_png.rf.f12285f93bb585acd36dcba02964a452.jpg
extracting: test/00000548_co_png.rf.9b0c3243f943ddc9df0f9a072bc96384.jpg
extracting: test/00000555_co_png.rf.0b25de64d8566a7d1c79fd88bac49ba7.jpg
extracting: test/00000568_co_png.rf.90471cf9b3a5d2866db1bcd973cd9170.jpg
extracting: test/00000582_co_png.rf.a3d88b90abd9b5dac4cf4d6fa08acd86.jpg
extracting: test/00000587_co_png.rf.bc3327b346b12082430a05e8d1637672.jpg
extracting: test/00000594_co_png.rf.58747fcff783f884d62a1db1a47b0207.jpg
extracting: test/_annotations.coco.json
creating: train/
extracting: train/00000267_co_png.rf.0ccd4864e9a4450b6f604eb37ecfc1a8.jpg
extracting: train/00000269_co_png.rf.d0f3aa521b0ece3b005a8cafcbe41a36.jpg
extracting: train/00000270_co_png.rf.e7e4b13aadb23f4ce3c486c3add85dc7.jpg
extracting: train/00000273_co_png.rf.0a3476fea12b9c20db035f0daa08d4d5.jpg
extracting: train/00000277_co_png.rf.8255348a3fdb5fcc0d76f1584b690fb1.jpg
extracting: train/00000290_co_png.rf.ebcb5a6bd2d88cbe0941e14ff02e8fad.jpg
extracting: train/00000291_co_png.rf.6e04348c411070715e00967d8ae7d845.jpg
extracting: train/00000292_co_png.rf.35687ffd365c58e8c0b275fac20f2e44.jpg
extracting: train/00000294_co_png.rf.c1d201e699a0af47b4270b9b1aaa4cd2.jpg
extracting: train/00000296_co_png.rf.a960c14460aa516564efb7261707b109.jpg
extracting: train/00000297_co_png.rf.66dfc34a1dc1504810f31973ba0b2039.jpg
extracting: train/00000301_co_png.rf.6f503504a6e4daead2fa1e17ff6a67f0.jpg
extracting: train/00000304_co_png.rf.52c5b554a01f65cf1c9ee446f7dcb407.jpg
extracting: train/00000305_co_png.rf.c52f550fd8fcf60471a4436603f119b4.jpg
extracting: train/00000306_co_png.rf.60e88cfc3de74534901c1b82c790f78c.jpg
extracting: train/00000310_co_png.rf.cef4b6e51c1846b02f5a06a789212d16.jpg
extracting: train/00000311_co_png.rf.3b016fdd28a95de9b4814858f6d263e1.jpg
extracting: train/00000316_co_png.rf.f1eee848a7ca04fe67ba1a49566bddf2.jpg
extracting: train/00000534_co_png.rf.7e10e8d76b52137ef9d47d990536cee1.jpg
extracting: train/00000536_co_png.rf.4f53094747facbc11e5f7740fb6d3819.jpg
extracting: train/00000537_co_png.rf.2ed5c658caf0031f7c43e1e7a31e2267.jpg
extracting: train/00000540_co_png.rf.9ba3b61fa463e58279231def4ce80fa4.jpg
extracting: train/00000541_co_png.rf.c4af049b34f5dae7ff3c4fbc22138d15.jpg
extracting: train/00000542_co_png.rf.dfc2ca18cd13840dbdcc4be7583111fb.jpg
extracting: train/00000543_co_png.rf.3473a6f96a923d1d1a106b32c8a0d4d4.jpg
extracting: train/00000544_co_png.rf.8ef793d4aadb9bee1ec9250d28cff02c.jpg
extracting: train/00000547_co_png.rf.cd892e14ced55238bf7a8570ad49c6b2.jpg
extracting: train/00000549_co_png.rf.53fff3921e14cd7c0a593c6906cfdee5.jpg
extracting: train/00000550_co_png.rf.ac5a63b5ced2c7d961bb06a884930418.jpg
extracting: train/00000551_co_png.rf.2314b153b556e1e5d12c62869df791f8.jpg
extracting: train/00000552_co_png.rf.a8834a205a545bb443ad2f44369108f9.jpg
extracting: train/00000553_co_png.rf.57e2994ae96b6c45bafe63979a6b1bdd.jpg
extracting: train/00000554_co_png.rf.379eec3504d0343d21d42fc3222af449.jpg
extracting: train/00000557_co_png.rf.12a65a5aa10632938ba93f7457237180.jpg
extracting: train/00000558_co_png.rf.e62112e71797b65428c44320ad52d663.jpg
extracting: train/00000559_co_png.rf.c2d2cdceaeada0e41ca77cc266ffbfde.jpg
extracting: train/00000561_co_png.rf.c0e100771651f5fb2d43e2f96a1df3fb.jpg
extracting: train/00000562_co_png.rf.d4be02e068051aa45f10df9e111cf0a5.jpg
extracting: train/00000563_co_png.rf.64b3e8bbfd738b2b3711e3db08180ea9.jpg
extracting: train/00000564_co_png.rf.28bd4e2b837ce01f6a2e059d76bb0944.jpg
extracting: train/00000565_co_png.rf.71f7d890fed6f3587911ee1e3975eb68.jpg
extracting: train/00000566_co_png.rf.101bcf8e64d7ef2547da136765c2ab1f.jpg
extracting: train/00000567_co_png.rf.47ce0b7ffba55aeb8a52ff83bc1e2ae3.jpg
extracting: train/00000569_co_png.rf.0c0ace952454065ae7f2c33da50cb728.jpg
extracting: train/00000570_co_png.rf.2a502639d1893e4f9ba82f2bcc89a07a.jpg
extracting: train/00000573_co_png.rf.7cd099696aaee46fe52329d61254503a.jpg
extracting: train/00000575_co_png.rf.55ff258f27c5aec0a7448e6effa7bb9e.jpg
extracting: train/00000577_co_png.rf.8ea72469282ffd2f139afc5026a8b4c1.jpg
extracting: train/00000578_co_png.rf.a0151c0270f688a092dab819a4ea94f4.jpg
extracting: train/00000579_co_png.rf.13a9bf95b4c7339a31d76374c081ab1e.jpg
extracting: train/00000580_co_png.rf.3e9df28a074589935c05d896579c9b96.jpg
extracting: train/00000585_co_png.rf.de5e53a2e5ea26ed775ce287281108eb.jpg
extracting: train/00000586_co_png.rf.6cc1033899d2be17e15b42012fbed22b.jpg
extracting: train/00000588_co_png.rf.a30758e75bc433110c1b483a037815b1.jpg
extracting: train/00000589_co_png.rf.fba98d43a62b53c8fe05be3893b206a3.jpg
extracting: train/00000590_co_png.rf.6dbe7d3aca042b9476156832909bbe28.jpg
extracting: train/00000592_co_png.rf.5c8e42d398468e481f135a5b9ef31881.jpg
extracting: train/00000593_co_png.rf.6e6555db9e7f9b3ab72d418afe241c6d.jpg
extracting: train/00000595_co_png.rf.e9dc5ef39cdeaa0a65af9948e2a67969.jpg
extracting: train/00000596_co_png.rf.c71749e1be314ac76cc6f9e15c90c6d8.jpg
extracting: train/00000597_co_png.rf.e1fa2a86627da23e6c2ffd01ba7a3553.jpg
extracting: train/00000598_co_png.rf.3fdcf67d894c7ed54ac2587952908873.jpg
extracting: train/00000599_co_png.rf.e68cdfbd175d412976011ce248b2316a.jpg
extracting: train/00000601_co_png.rf.f865ddfe62005641985545d30e495b16.jpg
extracting: train/00000602_co_png.rf.d5af4dec6f03c463fb279de3f13100dd.jpg
extracting: train/00000603_co_png.rf.500cad664275275b266b4e9e79d3c994.jpg
extracting: train/00000604_co_png.rf.91bafe4f00506b8e480ebef2e6bafa1d.jpg
extracting: train/00000605_co_png.rf.21779fa081156cfa6e403f25f7617c3e.jpg
extracting: train/00000607_co_png.rf.32cc474583d2955647e59ae24be01d6e.jpg
extracting: train/00000608_co_png.rf.2085593185e0f2679b8634371cb47236.jpg
extracting: train/00000609_co_png.rf.ca824722dea417a4ba6ad85ee17d8317.jpg
extracting: train/00000610_co_png.rf.46a0ccb5f0ee434b4ab84dd618f2ce43.jpg
extracting: train/00000611_co_png.rf.966d05f76017a690990cfa156d8df28c.jpg
extracting: train/00000612_co_png.rf.0c5125390ee327f33a481c3961ebd84a.jpg
extracting: train/00000613_co_png.rf.3c5475c2366b6eaeb0b240557455185e.jpg
extracting: train/00000614_co_png.rf.71b8ae596c297472e4e18a522afbbc99.jpg
extracting: train/00000615_co_png.rf.fd8ab599083cc0d49db5930bad0ce444.jpg
extracting: train/00000616_co_png.rf.2c06e199bd590f7ab2ccf5ec260f38ff.jpg
extracting: train/00000617_co_png.rf.a66f8027b85afff851c836f636027a6d.jpg
extracting: train/00000618_co_png.rf.64867f7f460a825c533bee3d5957d6af.jpg
extracting: train/00000619_co_png.rf.a2913ca56a6a805d875c0098fef547aa.jpg
extracting: train/00000620_co_png.rf.01e842a62572d946abd3b3b1e33ff006.jpg
extracting: train/00000621_co_png.rf.29745456f6c3a4724c1ce9a5ce44ab55.jpg
extracting: train/00000622_co_png.rf.5ff4bf4aee1040dace01b11c803b1347.jpg
extracting: train/00000623_co_png.rf.b72f15bec8ea2aa558a0db7fb50d4215.jpg
extracting: train/00000624_co_png.rf.e2cb7ad67e31d56347a59e5bed264e6d.jpg
extracting: train/00000625_co_png.rf.cd602a90557b624df33bae867eb031a7.jpg
extracting: train/00000626_co_png.rf.9f3d08394211564b466a144e2a34ead4.jpg
extracting: train/00000627_co_png.rf.a32c9c6149a9ff8d667d9f1f50f7d32e.jpg
extracting: train/00000628_co_png.rf.18cafc8b20326694e69add6444be4896.jpg
extracting: train/00000629_co_png.rf.b26b591c155cc028da4ef88c04d17bbc.jpg
extracting: train/00000630_co_png.rf.d9c9e9ddd9bcf68ec2d78bb957eeaa42.jpg
extracting: train/00000631_co_png.rf.39d37a44c52e5b236c391b3fa3b59d5f.jpg
extracting: train/00000632_co_png.rf.27f96770999e568075affd37058362bb.jpg
extracting: train/00000633_co_png.rf.560cbe0e0e609dc09f96dca4dc441b75.jpg
extracting: train/00000634_co_png.rf.2ff2ba8264106e05c040506fdb0f89cf.jpg
extracting: train/00000635_co_png.rf.26aa8aedb4600c85963ed8eb0b1ff806.jpg
extracting: train/00000636_co_png.rf.52beec6ef64d1b4cc8a8c80dba3714b3.jpg
extracting: train/00000637_co_png.rf.66281581f6d78a1823e27393c1715898.jpg
extracting: train/00000638_co_png.rf.c2ea7883861ce8f67fed04ecc6da6fc1.jpg
extracting: train/00000639_co_png.rf.d3111d6e7fb9e91e99542dfe6cd504cb.jpg
extracting: train/00000640_co_png.rf.ea61cde8b0d42989661251dc445580b0.jpg
extracting: train/00000641_co_png.rf.94c1a1610282127ad214bc692508e3d4.jpg
extracting: train/00000642_co_png.rf.d04bad534b8065b224ae10f024104a70.jpg
extracting: train/00000643_co_png.rf.301873a155ca88f20c039a9eaf2aa628.jpg
extracting: train/00000644_co_png.rf.e5bb4ac5170c35ed81e1ffd795aad026.jpg
extracting: train/00000645_co_png.rf.7c68ea8db8d994f49263f4a03b13eae1.jpg
extracting: train/00000646_co_png.rf.4d1528ec75bd4f4ac75cd5f2d19c9d2c.jpg
extracting: train/00000647_co_png.rf.621b32c73cb3de15ef649ed6482a0315.jpg
extracting: train/00000648_co_png.rf.d2eb67676f2ff3b36f05eb6bd5e1f558.jpg
extracting: train/00000649_co_png.rf.eb09f29d58f94e4ff5dbd61f2c0eb76c.jpg
extracting: train/00000650_co_png.rf.af78027f78d5f4982ac8d2d417926c96.jpg
extracting: train/00000651_co_png.rf.cd1bb79604b495528706f9d2a59aab32.jpg
extracting: train/00000652_co_png.rf.1a286d3113572b2acbb567cf7de28bc5.jpg
extracting: train/00000653_co_png.rf.772a5e851c0dd12c257d17d681016034.jpg
extracting: train/00000654_co_png.rf.d2fd5a1eb8bdf35a5f74f1f2d1d33845.jpg
extracting: train/00000655_co_png.rf.7911c9e96bf081258f0b8cf201d0c837.jpg
extracting: train/00000656_co_png.rf.6e234b6a44a060f72d5c93ca0f4d49d8.jpg
extracting: train/00000657_co_png.rf.7848747f48920308c197a3244db47e1b.jpg
extracting: train/00000658_co_png.rf.373ea2bb530d049dea863fc6d350bf89.jpg
extracting: train/00000659_co_png.rf.5a5e3bdff494e85265e14812c6371733.jpg
extracting: train/00000660_co_png.rf.fcf1a5d8dc8b4eaef27fc5e8a35df973.jpg
extracting: train/00000661_co_png.rf.4b5fe5062bf59db8944be575d36002b5.jpg
extracting: train/00000662_co_png.rf.aa8d3e144261ecdc70c335979e86f9cd.jpg
extracting: train/00000663_co_png.rf.79da21ec01ed3eeb774f691d67c96baf.jpg
extracting: train/00000664_co_png.rf.b6d76f675fb23f6d5e5c8ab69c175622.jpg
extracting: train/00000665_co_png.rf.69eb5ceb42d6e7ef5dc6f7ac2a21daab.jpg
extracting: train/00000666_co_png.rf.ad88041adb86086e395cc567d230dd67.jpg
extracting: train/00000667_co_png.rf.e0ff0c78e7d95c1d87fe982f53ff66ce.jpg
extracting: train/00000668_co_png.rf.80d8a3111f806b8502a74bac23ea9dfa.jpg
extracting: train/00000669_co_png.rf.1641ff92d29e2184c646860910b93a4f.jpg
extracting: train/00000670_co_png.rf.179e2c96647ac7548fc74afa6224cc93.jpg
extracting: train/00000671_co_png.rf.fd0c52df26c7eca2acc12d079a14fb3d.jpg
extracting: train/00000672_co_png.rf.acaf0cbd16b2ccee89fe851ef00c5c1e.jpg
extracting: train/00000673_co_png.rf.9993a1259c10c1a6863cd0b94dbab0e5.jpg
extracting: train/00000674_co_png.rf.f141beb272ba06a3b646df253a9b145d.jpg
extracting: train/00000675_co_png.rf.92704b802a0b439eb1c96d00b11b3ed9.jpg
extracting: train/00000676_co_png.rf.78d81b6b09a7be07a397af9229ba5e93.jpg
extracting: train/00000677_co_png.rf.b2e69ab428bf098664677b6a85a0884c.jpg
extracting: train/00000678_co_png.rf.e00ed743335d660882708ce613b52db5.jpg
extracting: train/00000679_co_png.rf.7d70b103821b4142e43f52f653950e3b.jpg
extracting: train/00000680_co_png.rf.45ded170ecfc327de476036027811614.jpg
extracting: train/00000681_co_png.rf.48a89166a2a7b441bbb9918278b250d6.jpg
extracting: train/00000682_co_png.rf.a7a4ba92849ad4b7c5ed0a4b29a43b27.jpg
extracting: train/00000683_co_png.rf.8636ea9d0abd45a72c03d77a135cced0.jpg
extracting: train/00000684_co_png.rf.76c6eabf7f74b76366bc439f38a49319.jpg
extracting: train/00000685_co_png.rf.0a00a521e80ef8211dbc81b0f49178fb.jpg
extracting: train/00000686_co_png.rf.d655e6558e9d50d696748a6f76b12380.jpg
extracting: train/00000687_co_png.rf.7e21cfd09d03cea3b21751479007ec88.jpg
extracting: train/00000688_co_png.rf.fc4b839f9e739ef9e038614b1123b58e.jpg
extracting: train/_annotations.coco.json
creating: valid/
extracting: valid/00000535_co_png.rf.4a5addf3a3ede9f3ae9f2c1c590046d7.jpg
extracting: valid/00000545_co_png.rf.b9644c032c02a8d17496cf4ff0f843c9.jpg
extracting: valid/00000546_co_png.rf.c08d5342b76d6c860e7996122ef970e5.jpg
extracting: valid/00000556_co_png.rf.591bb9fa52d3509e579e9c8ab25b9020.jpg
extracting: valid/00000571_co_png.rf.0393008d677d2f4f845ef360e8b5385c.jpg
extracting: valid/00000572_co_png.rf.d5873b9e22a29a46e2fb1915aea34566.jpg
extracting: valid/00000574_co_png.rf.f178af6d042313237b141123fe6b5efc.jpg
extracting: valid/00000576_co_png.rf.06bf1d628e9e5edee8ff0b3230a10a41.jpg
extracting: valid/00000583_co_png.rf.9b65de842574bcc4b2b2084b5df100a3.jpg
extracting: valid/00000591_co_png.rf.19fd54775f29e93d3f6df0553668d778.jpg
extracting: valid/_annotations.coco.json
Register & Visualize the dataset¶
In [ ]:
#Register to my custom dataset: train - 160, valid - 10, test - 40
register_coco_instances("my_dataset_train", {}, "/content/train/_annotations.coco.json", "/content/train")
register_coco_instances("my_dataset_val", {}, "/content/valid/_annotations.coco.json", "/content/valid")
register_coco_instances("my_dataset_test", {}, "/content/test/_annotations.coco.json", "/content/test")
In [ ]:
#visualize training data
my_dataset_train_metadata = MetadataCatalog.get("my_dataset_train")
dataset_dicts = DatasetCatalog.get("my_dataset_train")
for d in random.sample(dataset_dicts, 3):
img = cv2.imread(d["file_name"])
visualizer = Visualizer(img[:, :, ::-1], metadata=my_dataset_train_metadata, scale=0.5)
vis = visualizer.draw_dataset_dict(d)
cv2_imshow(vis.get_image()[:, :, ::-1])
WARNING [09/28 16:59:14 d2.data.datasets.coco]: Category ids in annotations are not in [1, #categories]! We'll apply a mapping for you. [09/28 16:59:14 d2.data.datasets.coco]: Loaded 150 images in COCO format from /content/train/_annotations.coco.json
Train my custom detector¶
In [ ]:
#We are importing our own Trainer Module here to use the COCO validation evaluation during training. Otherwise no validation eval occurs.
class CocoTrainer(DefaultTrainer):
@classmethod
def build_evaluator(cls, cfg, dataset_name, output_folder=None):
if output_folder is None:
os.makedirs("coco_eval", exist_ok=True)
output_folder = "coco_eval"
return COCOEvaluator(dataset_name, cfg, False, output_folder)
In [ ]:
cfg = get_cfg() # cfgNode
cfg.merge_from_file(model_zoo.get_config_file("COCO-Detection/faster_rcnn_X_101_32x8d_FPN_3x.yaml")) # get the model configuration
cfg.DATASETS.TRAIN = ("my_dataset_train",) # set the train dataset
cfg.DATASETS.TEST = ("my_dataset_val",) # set the validation dataset
cfg.DATALOADER.NUM_WORKERS = 4 # num of workers is the computer num of cores
cfg.MODEL.WEIGHTS = model_zoo.get_checkpoint_url("COCO-Detection/faster_rcnn_X_101_32x8d_FPN_3x.yaml") # Let training initialize from model zoo
cfg.SOLVER.IMS_PER_BATCH = 4 # default definition
cfg.SOLVER.BASE_LR = 0.001 # default definition
cfg.SOLVER.WARMUP_ITERS = 1000
cfg.SOLVER.MAX_ITER = 1000 # epochs number
cfg.SOLVER.STEPS = (1000, 1500)
cfg.SOLVER.GAMMA = 0.05
cfg.MODEL.ROI_HEADS.BATCH_SIZE_PER_IMAGE = 64
cfg.MODEL.ROI_HEADS.NUM_CLASSES = 2 #number of classes + 1
cfg.TEST.EVAL_PERIOD = 500
os.makedirs(cfg.OUTPUT_DIR, exist_ok=True)
trainer = CocoTrainer(cfg) # create the trainer
trainer.resume_or_load(resume=False)
trainer.train() # train
[09/04 18:28:52 d2.engine.defaults]: Model: GeneralizedRCNN( (backbone): FPN( (fpn_lateral2): Conv2d(256, 256, kernel_size=(1, 1), stride=(1, 1)) (fpn_output2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)) (fpn_lateral3): Conv2d(512, 256, kernel_size=(1, 1), stride=(1, 1)) (fpn_output3): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)) (fpn_lateral4): Conv2d(1024, 256, kernel_size=(1, 1), stride=(1, 1)) (fpn_output4): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)) (fpn_lateral5): Conv2d(2048, 256, kernel_size=(1, 1), stride=(1, 1)) (fpn_output5): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)) (top_block): LastLevelMaxPool() (bottom_up): ResNet( (stem): BasicStem( (conv1): Conv2d( 3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05) ) ) (res2): Sequential( (0): BottleneckBlock( (shortcut): Conv2d( 64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05) ) (conv1): Conv2d( 64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05) ) (conv2): Conv2d( 256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=32, bias=False (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05) ) (conv3): Conv2d( 256, 256, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05) ) ) (1): BottleneckBlock( (conv1): Conv2d( 256, 256, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05) ) (conv2): Conv2d( 256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=32, bias=False (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05) ) (conv3): Conv2d( 256, 256, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05) ) ) (2): BottleneckBlock( (conv1): Conv2d( 256, 256, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05) ) (conv2): Conv2d( 256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=32, bias=False (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05) ) (conv3): Conv2d( 256, 256, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05) ) ) ) (res3): Sequential( (0): BottleneckBlock( (shortcut): Conv2d( 256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05) ) (conv1): Conv2d( 256, 512, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05) ) (conv2): Conv2d( 512, 512, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), groups=32, bias=False (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05) ) (conv3): Conv2d( 512, 512, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05) ) ) (1): BottleneckBlock( (conv1): Conv2d( 512, 512, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05) ) (conv2): Conv2d( 512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=32, bias=False (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05) ) (conv3): Conv2d( 512, 512, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05) ) ) (2): BottleneckBlock( (conv1): Conv2d( 512, 512, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05) ) (conv2): Conv2d( 512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=32, bias=False (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05) ) (conv3): Conv2d( 512, 512, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05) ) ) (3): BottleneckBlock( (conv1): Conv2d( 512, 512, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05) ) (conv2): Conv2d( 512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=32, bias=False (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05) ) (conv3): Conv2d( 512, 512, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05) ) ) ) (res4): Sequential( (0): BottleneckBlock( (shortcut): Conv2d( 512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False (norm): FrozenBatchNorm2d(num_features=1024, eps=1e-05) ) (conv1): Conv2d( 512, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=1024, eps=1e-05) ) (conv2): Conv2d( 1024, 1024, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), groups=32, bias=False (norm): FrozenBatchNorm2d(num_features=1024, eps=1e-05) ) (conv3): Conv2d( 1024, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=1024, eps=1e-05) ) ) (1): BottleneckBlock( (conv1): Conv2d( 1024, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=1024, eps=1e-05) ) (conv2): Conv2d( 1024, 1024, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=32, bias=False (norm): FrozenBatchNorm2d(num_features=1024, eps=1e-05) ) (conv3): Conv2d( 1024, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=1024, eps=1e-05) ) ) (2): BottleneckBlock( (conv1): Conv2d( 1024, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=1024, eps=1e-05) ) (conv2): Conv2d( 1024, 1024, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=32, bias=False (norm): FrozenBatchNorm2d(num_features=1024, eps=1e-05) ) (conv3): Conv2d( 1024, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=1024, eps=1e-05) ) ) (3): BottleneckBlock( (conv1): Conv2d( 1024, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=1024, eps=1e-05) ) (conv2): Conv2d( 1024, 1024, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=32, bias=False (norm): FrozenBatchNorm2d(num_features=1024, eps=1e-05) ) (conv3): Conv2d( 1024, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=1024, eps=1e-05) ) ) (4): BottleneckBlock( (conv1): Conv2d( 1024, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=1024, eps=1e-05) ) (conv2): Conv2d( 1024, 1024, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=32, bias=False (norm): FrozenBatchNorm2d(num_features=1024, eps=1e-05) ) (conv3): Conv2d( 1024, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=1024, eps=1e-05) ) ) (5): BottleneckBlock( (conv1): Conv2d( 1024, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=1024, eps=1e-05) ) (conv2): Conv2d( 1024, 1024, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=32, bias=False (norm): FrozenBatchNorm2d(num_features=1024, eps=1e-05) ) (conv3): Conv2d( 1024, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=1024, eps=1e-05) ) ) (6): BottleneckBlock( (conv1): Conv2d( 1024, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=1024, eps=1e-05) ) (conv2): Conv2d( 1024, 1024, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=32, bias=False (norm): FrozenBatchNorm2d(num_features=1024, eps=1e-05) ) (conv3): Conv2d( 1024, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=1024, eps=1e-05) ) ) (7): BottleneckBlock( (conv1): Conv2d( 1024, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=1024, eps=1e-05) ) (conv2): Conv2d( 1024, 1024, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=32, bias=False (norm): FrozenBatchNorm2d(num_features=1024, eps=1e-05) ) (conv3): Conv2d( 1024, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=1024, eps=1e-05) ) ) (8): BottleneckBlock( (conv1): Conv2d( 1024, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=1024, eps=1e-05) ) (conv2): Conv2d( 1024, 1024, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=32, bias=False (norm): FrozenBatchNorm2d(num_features=1024, eps=1e-05) ) (conv3): Conv2d( 1024, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=1024, eps=1e-05) ) ) (9): BottleneckBlock( (conv1): Conv2d( 1024, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=1024, eps=1e-05) ) (conv2): Conv2d( 1024, 1024, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=32, bias=False (norm): FrozenBatchNorm2d(num_features=1024, eps=1e-05) ) (conv3): Conv2d( 1024, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=1024, eps=1e-05) ) ) (10): BottleneckBlock( (conv1): Conv2d( 1024, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=1024, eps=1e-05) ) (conv2): Conv2d( 1024, 1024, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=32, bias=False (norm): FrozenBatchNorm2d(num_features=1024, eps=1e-05) ) (conv3): Conv2d( 1024, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=1024, eps=1e-05) ) ) (11): BottleneckBlock( (conv1): Conv2d( 1024, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=1024, eps=1e-05) ) (conv2): Conv2d( 1024, 1024, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=32, bias=False (norm): FrozenBatchNorm2d(num_features=1024, eps=1e-05) ) (conv3): Conv2d( 1024, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=1024, eps=1e-05) ) ) (12): BottleneckBlock( (conv1): Conv2d( 1024, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=1024, eps=1e-05) ) (conv2): Conv2d( 1024, 1024, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=32, bias=False (norm): FrozenBatchNorm2d(num_features=1024, eps=1e-05) ) (conv3): Conv2d( 1024, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=1024, eps=1e-05) ) ) (13): BottleneckBlock( (conv1): Conv2d( 1024, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=1024, eps=1e-05) ) (conv2): Conv2d( 1024, 1024, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=32, bias=False (norm): FrozenBatchNorm2d(num_features=1024, eps=1e-05) ) (conv3): Conv2d( 1024, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=1024, eps=1e-05) ) ) (14): BottleneckBlock( (conv1): Conv2d( 1024, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=1024, eps=1e-05) ) (conv2): Conv2d( 1024, 1024, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=32, bias=False (norm): FrozenBatchNorm2d(num_features=1024, eps=1e-05) ) (conv3): Conv2d( 1024, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=1024, eps=1e-05) ) ) (15): BottleneckBlock( (conv1): Conv2d( 1024, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=1024, eps=1e-05) ) (conv2): Conv2d( 1024, 1024, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=32, bias=False (norm): FrozenBatchNorm2d(num_features=1024, eps=1e-05) ) (conv3): Conv2d( 1024, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=1024, eps=1e-05) ) ) (16): BottleneckBlock( (conv1): Conv2d( 1024, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=1024, eps=1e-05) ) (conv2): Conv2d( 1024, 1024, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=32, bias=False (norm): FrozenBatchNorm2d(num_features=1024, eps=1e-05) ) (conv3): Conv2d( 1024, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=1024, eps=1e-05) ) ) (17): BottleneckBlock( (conv1): Conv2d( 1024, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=1024, eps=1e-05) ) (conv2): Conv2d( 1024, 1024, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=32, bias=False (norm): FrozenBatchNorm2d(num_features=1024, eps=1e-05) ) (conv3): Conv2d( 1024, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=1024, eps=1e-05) ) ) (18): BottleneckBlock( (conv1): Conv2d( 1024, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=1024, eps=1e-05) ) (conv2): Conv2d( 1024, 1024, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=32, bias=False (norm): FrozenBatchNorm2d(num_features=1024, eps=1e-05) ) (conv3): Conv2d( 1024, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=1024, eps=1e-05) ) ) (19): BottleneckBlock( (conv1): Conv2d( 1024, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=1024, eps=1e-05) ) (conv2): Conv2d( 1024, 1024, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=32, bias=False (norm): FrozenBatchNorm2d(num_features=1024, eps=1e-05) ) (conv3): Conv2d( 1024, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=1024, eps=1e-05) ) ) (20): BottleneckBlock( (conv1): Conv2d( 1024, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=1024, eps=1e-05) ) (conv2): Conv2d( 1024, 1024, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=32, bias=False (norm): FrozenBatchNorm2d(num_features=1024, eps=1e-05) ) (conv3): Conv2d( 1024, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=1024, eps=1e-05) ) ) (21): BottleneckBlock( (conv1): Conv2d( 1024, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=1024, eps=1e-05) ) (conv2): Conv2d( 1024, 1024, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=32, bias=False (norm): FrozenBatchNorm2d(num_features=1024, eps=1e-05) ) (conv3): Conv2d( 1024, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=1024, eps=1e-05) ) ) (22): BottleneckBlock( (conv1): Conv2d( 1024, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=1024, eps=1e-05) ) (conv2): Conv2d( 1024, 1024, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=32, bias=False (norm): FrozenBatchNorm2d(num_features=1024, eps=1e-05) ) (conv3): Conv2d( 1024, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=1024, eps=1e-05) ) ) ) (res5): Sequential( (0): BottleneckBlock( (shortcut): Conv2d( 1024, 2048, kernel_size=(1, 1), stride=(2, 2), bias=False (norm): FrozenBatchNorm2d(num_features=2048, eps=1e-05) ) (conv1): Conv2d( 1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=2048, eps=1e-05) ) (conv2): Conv2d( 2048, 2048, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), groups=32, bias=False (norm): FrozenBatchNorm2d(num_features=2048, eps=1e-05) ) (conv3): Conv2d( 2048, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=2048, eps=1e-05) ) ) (1): BottleneckBlock( (conv1): Conv2d( 2048, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=2048, eps=1e-05) ) (conv2): Conv2d( 2048, 2048, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=32, bias=False (norm): FrozenBatchNorm2d(num_features=2048, eps=1e-05) ) (conv3): Conv2d( 2048, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=2048, eps=1e-05) ) ) (2): BottleneckBlock( (conv1): Conv2d( 2048, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=2048, eps=1e-05) ) (conv2): Conv2d( 2048, 2048, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=32, bias=False (norm): FrozenBatchNorm2d(num_features=2048, eps=1e-05) ) (conv3): Conv2d( 2048, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False (norm): FrozenBatchNorm2d(num_features=2048, eps=1e-05) ) ) ) ) ) (proposal_generator): RPN( (anchor_generator): DefaultAnchorGenerator( (cell_anchors): BufferList() ) (rpn_head): StandardRPNHead( (conv): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)) (objectness_logits): Conv2d(256, 3, kernel_size=(1, 1), stride=(1, 1)) (anchor_deltas): Conv2d(256, 12, kernel_size=(1, 1), stride=(1, 1)) ) ) (roi_heads): StandardROIHeads( (box_pooler): ROIPooler( (level_poolers): ModuleList( (0): ROIAlign(output_size=(7, 7), spatial_scale=0.25, sampling_ratio=0, aligned=True) (1): ROIAlign(output_size=(7, 7), spatial_scale=0.125, sampling_ratio=0, aligned=True) (2): ROIAlign(output_size=(7, 7), spatial_scale=0.0625, sampling_ratio=0, aligned=True) (3): ROIAlign(output_size=(7, 7), spatial_scale=0.03125, sampling_ratio=0, aligned=True) ) ) (box_head): FastRCNNConvFCHead( (fc1): Linear(in_features=12544, out_features=1024, bias=True) (fc2): Linear(in_features=1024, out_features=1024, bias=True) ) (box_predictor): FastRCNNOutputLayers( (cls_score): Linear(in_features=1024, out_features=3, bias=True) (bbox_pred): Linear(in_features=1024, out_features=8, bias=True) ) ) ) WARNING [09/04 18:28:52 d2.data.datasets.coco]: Category ids in annotations are not in [1, #categories]! We'll apply a mapping for you. [09/04 18:28:52 d2.data.datasets.coco]: Loaded 150 images in COCO format from /content/train/_annotations.coco.json [09/04 18:28:52 d2.data.build]: Removed 4 images with no usable annotations. 146 images left. [09/04 18:28:52 d2.data.common]: Serializing 146 elements to byte tensors and concatenating them all ... [09/04 18:28:52 d2.data.common]: Serialized dataset takes 0.05 MiB [09/04 18:28:52 d2.data.detection_utils]: TransformGens used in training: [ResizeShortestEdge(short_edge_length=(640, 672, 704, 736, 768, 800), max_size=1333, sample_style='choice'), RandomFlip()] [09/04 18:28:52 d2.data.build]: Using training sampler TrainingSampler
WARNING:fvcore.common.checkpoint:Skip loading parameter 'roi_heads.box_predictor.cls_score.weight' to the model due to incompatible shapes: (81, 1024) in the checkpoint but (3, 1024) in the model! You might want to double check if this is expected. WARNING:fvcore.common.checkpoint:Skip loading parameter 'roi_heads.box_predictor.cls_score.bias' to the model due to incompatible shapes: (81,) in the checkpoint but (3,) in the model! You might want to double check if this is expected. WARNING:fvcore.common.checkpoint:Skip loading parameter 'roi_heads.box_predictor.bbox_pred.weight' to the model due to incompatible shapes: (320, 1024) in the checkpoint but (8, 1024) in the model! You might want to double check if this is expected. WARNING:fvcore.common.checkpoint:Skip loading parameter 'roi_heads.box_predictor.bbox_pred.bias' to the model due to incompatible shapes: (320,) in the checkpoint but (8,) in the model! You might want to double check if this is expected. WARNING:fvcore.common.checkpoint:Some model parameters or buffers are not found in the checkpoint: roi_heads.box_predictor.bbox_pred.{bias, weight} roi_heads.box_predictor.cls_score.{bias, weight}
[09/04 18:28:52 d2.engine.train_loop]: Starting training from iteration 0 [09/04 18:29:32 d2.utils.events]: eta: 0:33:35 iter: 19 total_loss: 1.718 loss_cls: 0.942 loss_box_reg: 0.466 loss_rpn_cls: 0.200 loss_rpn_loc: 0.021 time: 2.0451 data_time: 0.0200 lr: 0.000020 max_mem: 8426M [09/04 18:30:16 d2.utils.events]: eta: 0:34:11 iter: 39 total_loss: 1.505 loss_cls: 0.748 loss_box_reg: 0.518 loss_rpn_cls: 0.161 loss_rpn_loc: 0.015 time: 2.1252 data_time: 0.0093 lr: 0.000040 max_mem: 8426M [09/04 18:30:59 d2.utils.events]: eta: 0:33:38 iter: 59 total_loss: 1.408 loss_cls: 0.596 loss_box_reg: 0.647 loss_rpn_cls: 0.094 loss_rpn_loc: 0.018 time: 2.1330 data_time: 0.0102 lr: 0.000060 max_mem: 8426M [09/04 18:31:45 d2.utils.events]: eta: 0:33:16 iter: 79 total_loss: 1.323 loss_cls: 0.509 loss_box_reg: 0.731 loss_rpn_cls: 0.025 loss_rpn_loc: 0.017 time: 2.1658 data_time: 0.0092 lr: 0.000080 max_mem: 8426M [09/04 18:32:30 d2.utils.events]: eta: 0:32:46 iter: 99 total_loss: 1.216 loss_cls: 0.457 loss_box_reg: 0.715 loss_rpn_cls: 0.032 loss_rpn_loc: 0.017 time: 2.1820 data_time: 0.0087 lr: 0.000100 max_mem: 8426M [09/04 18:33:16 d2.utils.events]: eta: 0:32:27 iter: 119 total_loss: 1.156 loss_cls: 0.394 loss_box_reg: 0.739 loss_rpn_cls: 0.021 loss_rpn_loc: 0.015 time: 2.2022 data_time: 0.0100 lr: 0.000120 max_mem: 8426M [09/04 18:34:00 d2.utils.events]: eta: 0:31:55 iter: 139 total_loss: 1.143 loss_cls: 0.344 loss_box_reg: 0.719 loss_rpn_cls: 0.022 loss_rpn_loc: 0.018 time: 2.2077 data_time: 0.0094 lr: 0.000140 max_mem: 8426M [09/04 18:34:45 d2.utils.events]: eta: 0:31:12 iter: 159 total_loss: 1.059 loss_cls: 0.292 loss_box_reg: 0.728 loss_rpn_cls: 0.013 loss_rpn_loc: 0.016 time: 2.2136 data_time: 0.0090 lr: 0.000160 max_mem: 8426M [09/04 18:35:29 d2.utils.events]: eta: 0:30:23 iter: 179 total_loss: 1.027 loss_cls: 0.268 loss_box_reg: 0.716 loss_rpn_cls: 0.019 loss_rpn_loc: 0.017 time: 2.2094 data_time: 0.0101 lr: 0.000180 max_mem: 8426M [09/04 18:36:14 d2.utils.events]: eta: 0:29:43 iter: 199 total_loss: 0.897 loss_cls: 0.212 loss_box_reg: 0.667 loss_rpn_cls: 0.012 loss_rpn_loc: 0.014 time: 2.2147 data_time: 0.0103 lr: 0.000200 max_mem: 8426M [09/04 18:37:00 d2.utils.events]: eta: 0:29:02 iter: 219 total_loss: 0.942 loss_cls: 0.249 loss_box_reg: 0.642 loss_rpn_cls: 0.014 loss_rpn_loc: 0.016 time: 2.2214 data_time: 0.0090 lr: 0.000220 max_mem: 8426M [09/04 18:37:45 d2.utils.events]: eta: 0:28:16 iter: 239 total_loss: 0.809 loss_cls: 0.203 loss_box_reg: 0.585 loss_rpn_cls: 0.011 loss_rpn_loc: 0.015 time: 2.2236 data_time: 0.0088 lr: 0.000240 max_mem: 8426M [09/04 18:38:30 d2.utils.events]: eta: 0:27:31 iter: 259 total_loss: 0.755 loss_cls: 0.188 loss_box_reg: 0.512 loss_rpn_cls: 0.011 loss_rpn_loc: 0.017 time: 2.2241 data_time: 0.0093 lr: 0.000260 max_mem: 8426M [09/04 18:39:13 d2.utils.events]: eta: 0:26:44 iter: 279 total_loss: 0.733 loss_cls: 0.157 loss_box_reg: 0.541 loss_rpn_cls: 0.006 loss_rpn_loc: 0.015 time: 2.2206 data_time: 0.0094 lr: 0.000280 max_mem: 8426M [09/04 18:39:59 d2.utils.events]: eta: 0:26:03 iter: 299 total_loss: 0.761 loss_cls: 0.184 loss_box_reg: 0.543 loss_rpn_cls: 0.008 loss_rpn_loc: 0.019 time: 2.2264 data_time: 0.0096 lr: 0.000300 max_mem: 8426M [09/04 18:40:46 d2.utils.events]: eta: 0:25:22 iter: 319 total_loss: 0.652 loss_cls: 0.129 loss_box_reg: 0.459 loss_rpn_cls: 0.005 loss_rpn_loc: 0.009 time: 2.2320 data_time: 0.0096 lr: 0.000320 max_mem: 8426M [09/04 18:41:31 d2.utils.events]: eta: 0:24:39 iter: 339 total_loss: 0.672 loss_cls: 0.163 loss_box_reg: 0.479 loss_rpn_cls: 0.006 loss_rpn_loc: 0.015 time: 2.2336 data_time: 0.0094 lr: 0.000340 max_mem: 8426M [09/04 18:42:15 d2.utils.events]: eta: 0:23:53 iter: 359 total_loss: 0.667 loss_cls: 0.130 loss_box_reg: 0.506 loss_rpn_cls: 0.004 loss_rpn_loc: 0.015 time: 2.2331 data_time: 0.0090 lr: 0.000360 max_mem: 8426M [09/04 18:43:02 d2.utils.events]: eta: 0:23:11 iter: 379 total_loss: 0.571 loss_cls: 0.105 loss_box_reg: 0.434 loss_rpn_cls: 0.004 loss_rpn_loc: 0.009 time: 2.2396 data_time: 0.0090 lr: 0.000380 max_mem: 8426M [09/04 18:43:48 d2.utils.events]: eta: 0:22:27 iter: 399 total_loss: 0.635 loss_cls: 0.120 loss_box_reg: 0.494 loss_rpn_cls: 0.005 loss_rpn_loc: 0.013 time: 2.2410 data_time: 0.0094 lr: 0.000400 max_mem: 8426M [09/04 18:44:33 d2.utils.events]: eta: 0:21:43 iter: 419 total_loss: 0.591 loss_cls: 0.113 loss_box_reg: 0.446 loss_rpn_cls: 0.003 loss_rpn_loc: 0.015 time: 2.2411 data_time: 0.0091 lr: 0.000420 max_mem: 8426M [09/04 18:45:17 d2.utils.events]: eta: 0:20:58 iter: 439 total_loss: 0.546 loss_cls: 0.126 loss_box_reg: 0.413 loss_rpn_cls: 0.003 loss_rpn_loc: 0.011 time: 2.2405 data_time: 0.0090 lr: 0.000440 max_mem: 8426M [09/04 18:46:03 d2.utils.events]: eta: 0:20:14 iter: 459 total_loss: 0.552 loss_cls: 0.103 loss_box_reg: 0.422 loss_rpn_cls: 0.004 loss_rpn_loc: 0.013 time: 2.2423 data_time: 0.0104 lr: 0.000460 max_mem: 8426M [09/04 18:46:48 d2.utils.events]: eta: 0:19:29 iter: 479 total_loss: 0.540 loss_cls: 0.097 loss_box_reg: 0.408 loss_rpn_cls: 0.002 loss_rpn_loc: 0.010 time: 2.2423 data_time: 0.0090 lr: 0.000480 max_mem: 8426M WARNING [09/04 18:47:32 d2.data.datasets.coco]: Category ids in annotations are not in [1, #categories]! We'll apply a mapping for you. [09/04 18:47:32 d2.data.datasets.coco]: Loaded 10 images in COCO format from /content/valid/_annotations.coco.json [09/04 18:47:32 d2.data.common]: Serializing 10 elements to byte tensors and concatenating them all ... [09/04 18:47:32 d2.data.common]: Serialized dataset takes 0.00 MiB [09/04 18:47:32 d2.evaluation.evaluator]: Start inference on 10 images [09/04 18:47:35 d2.evaluation.evaluator]: Total inference time: 0:00:01.089128 (0.217826 s / img per device, on 1 devices) [09/04 18:47:35 d2.evaluation.evaluator]: Total inference pure compute time: 0:00:01 (0.201924 s / img per device, on 1 devices) [09/04 18:47:35 d2.evaluation.coco_evaluation]: Preparing results for COCO format ... [09/04 18:47:35 d2.evaluation.coco_evaluation]: Saving results to coco_eval/coco_instances_results.json [09/04 18:47:35 d2.evaluation.coco_evaluation]: Evaluating predictions ... Loading and preparing results... DONE (t=0.00s) creating index... index created! Running per image evaluation... Evaluate annotation type *bbox* DONE (t=0.01s). Accumulating evaluation results... DONE (t=0.01s). Average Precision (AP) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = 0.413 Average Precision (AP) @[ IoU=0.50 | area= all | maxDets=100 ] = 0.640 Average Precision (AP) @[ IoU=0.75 | area= all | maxDets=100 ] = 0.486 Average Precision (AP) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.414 Average Precision (AP) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = -1.000 Average Precision (AP) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = -1.000 Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets= 1 ] = 0.414 Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets= 10 ] = 0.471 Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = 0.471 Average Recall (AR) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.471 Average Recall (AR) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = -1.000 Average Recall (AR) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = -1.000 [09/04 18:47:35 d2.evaluation.coco_evaluation]: Evaluation results for bbox: | AP | AP50 | AP75 | APs | APm | APl | |:------:|:------:|:------:|:------:|:-----:|:-----:| | 41.276 | 63.976 | 48.639 | 41.355 | nan | nan | [09/04 18:47:35 d2.evaluation.coco_evaluation]: Note that some metrics cannot be computed. [09/04 18:47:35 d2.evaluation.coco_evaluation]: Per-category bbox AP: | category | AP | category | AP | |:-----------|:-----|:-----------|:-------| | Car | nan | Car | 41.276 | [09/04 18:47:35 d2.engine.defaults]: Evaluation results for my_dataset_val in csv format: [09/04 18:47:35 d2.evaluation.testing]: copypaste: Task: bbox [09/04 18:47:35 d2.evaluation.testing]: copypaste: AP,AP50,AP75,APs,APm,APl [09/04 18:47:35 d2.evaluation.testing]: copypaste: 41.2757,63.9764,48.6386,41.3549,nan,nan [09/04 18:47:35 d2.utils.events]: eta: 0:18:43 iter: 499 total_loss: 0.523 loss_cls: 0.102 loss_box_reg: 0.409 loss_rpn_cls: 0.003 loss_rpn_loc: 0.013 time: 2.2406 data_time: 0.0090 lr: 0.000500 max_mem: 8426M [09/04 18:48:19 d2.utils.events]: eta: 0:17:58 iter: 519 total_loss: 0.508 loss_cls: 0.109 loss_box_reg: 0.393 loss_rpn_cls: 0.002 loss_rpn_loc: 0.010 time: 2.2404 data_time: 0.0093 lr: 0.000519 max_mem: 8426M [09/04 18:49:05 d2.utils.events]: eta: 0:17:14 iter: 539 total_loss: 0.523 loss_cls: 0.110 loss_box_reg: 0.366 loss_rpn_cls: 0.003 loss_rpn_loc: 0.011 time: 2.2414 data_time: 0.0089 lr: 0.000539 max_mem: 8426M [09/04 18:49:50 d2.utils.events]: eta: 0:16:30 iter: 559 total_loss: 0.504 loss_cls: 0.109 loss_box_reg: 0.373 loss_rpn_cls: 0.002 loss_rpn_loc: 0.011 time: 2.2420 data_time: 0.0094 lr: 0.000559 max_mem: 8426M [09/04 18:50:35 d2.utils.events]: eta: 0:15:45 iter: 579 total_loss: 0.520 loss_cls: 0.102 loss_box_reg: 0.399 loss_rpn_cls: 0.001 loss_rpn_loc: 0.011 time: 2.2429 data_time: 0.0089 lr: 0.000579 max_mem: 8426M [09/04 18:51:22 d2.utils.events]: eta: 0:15:00 iter: 599 total_loss: 0.456 loss_cls: 0.091 loss_box_reg: 0.371 loss_rpn_cls: 0.001 loss_rpn_loc: 0.008 time: 2.2454 data_time: 0.0092 lr: 0.000599 max_mem: 8426M [09/04 18:52:09 d2.utils.events]: eta: 0:14:16 iter: 619 total_loss: 0.491 loss_cls: 0.095 loss_box_reg: 0.381 loss_rpn_cls: 0.002 loss_rpn_loc: 0.012 time: 2.2490 data_time: 0.0088 lr: 0.000619 max_mem: 8426M [09/04 18:52:54 d2.utils.events]: eta: 0:13:31 iter: 639 total_loss: 0.459 loss_cls: 0.091 loss_box_reg: 0.341 loss_rpn_cls: 0.001 loss_rpn_loc: 0.010 time: 2.2489 data_time: 0.0095 lr: 0.000639 max_mem: 8426M [09/04 18:53:38 d2.utils.events]: eta: 0:12:46 iter: 659 total_loss: 0.468 loss_cls: 0.087 loss_box_reg: 0.370 loss_rpn_cls: 0.002 loss_rpn_loc: 0.012 time: 2.2480 data_time: 0.0088 lr: 0.000659 max_mem: 8426M [09/04 18:54:25 d2.utils.events]: eta: 0:12:03 iter: 679 total_loss: 0.483 loss_cls: 0.084 loss_box_reg: 0.367 loss_rpn_cls: 0.001 loss_rpn_loc: 0.011 time: 2.2502 data_time: 0.0098 lr: 0.000679 max_mem: 8426M [09/04 18:55:09 d2.utils.events]: eta: 0:11:17 iter: 699 total_loss: 0.402 loss_cls: 0.073 loss_box_reg: 0.314 loss_rpn_cls: 0.001 loss_rpn_loc: 0.010 time: 2.2492 data_time: 0.0091 lr: 0.000699 max_mem: 8426M [09/04 18:55:55 d2.utils.events]: eta: 0:10:33 iter: 719 total_loss: 0.416 loss_cls: 0.085 loss_box_reg: 0.307 loss_rpn_cls: 0.001 loss_rpn_loc: 0.010 time: 2.2513 data_time: 0.0098 lr: 0.000719 max_mem: 8426M [09/04 18:56:41 d2.utils.events]: eta: 0:09:48 iter: 739 total_loss: 0.425 loss_cls: 0.069 loss_box_reg: 0.336 loss_rpn_cls: 0.001 loss_rpn_loc: 0.010 time: 2.2526 data_time: 0.0087 lr: 0.000739 max_mem: 8426M [09/04 18:57:27 d2.utils.events]: eta: 0:09:03 iter: 759 total_loss: 0.457 loss_cls: 0.087 loss_box_reg: 0.355 loss_rpn_cls: 0.001 loss_rpn_loc: 0.012 time: 2.2533 data_time: 0.0086 lr: 0.000759 max_mem: 8426M [09/04 18:58:11 d2.utils.events]: eta: 0:08:18 iter: 779 total_loss: 0.401 loss_cls: 0.070 loss_box_reg: 0.304 loss_rpn_cls: 0.001 loss_rpn_loc: 0.007 time: 2.2524 data_time: 0.0094 lr: 0.000779 max_mem: 8426M [09/04 18:58:56 d2.utils.events]: eta: 0:07:33 iter: 799 total_loss: 0.394 loss_cls: 0.068 loss_box_reg: 0.311 loss_rpn_cls: 0.001 loss_rpn_loc: 0.011 time: 2.2514 data_time: 0.0092 lr: 0.000799 max_mem: 8426M [09/04 18:59:40 d2.utils.events]: eta: 0:06:47 iter: 819 total_loss: 0.401 loss_cls: 0.065 loss_box_reg: 0.322 loss_rpn_cls: 0.000 loss_rpn_loc: 0.009 time: 2.2504 data_time: 0.0085 lr: 0.000819 max_mem: 8426M [09/04 19:00:26 d2.utils.events]: eta: 0:06:02 iter: 839 total_loss: 0.397 loss_cls: 0.071 loss_box_reg: 0.317 loss_rpn_cls: 0.001 loss_rpn_loc: 0.010 time: 2.2513 data_time: 0.0093 lr: 0.000839 max_mem: 8426M [09/04 19:01:11 d2.utils.events]: eta: 0:05:17 iter: 859 total_loss: 0.350 loss_cls: 0.062 loss_box_reg: 0.276 loss_rpn_cls: 0.001 loss_rpn_loc: 0.009 time: 2.2515 data_time: 0.0087 lr: 0.000859 max_mem: 8426M [09/04 19:01:57 d2.utils.events]: eta: 0:04:32 iter: 879 total_loss: 0.392 loss_cls: 0.059 loss_box_reg: 0.316 loss_rpn_cls: 0.001 loss_rpn_loc: 0.009 time: 2.2527 data_time: 0.0093 lr: 0.000879 max_mem: 8426M [09/04 19:02:40 d2.utils.events]: eta: 0:03:47 iter: 899 total_loss: 0.393 loss_cls: 0.061 loss_box_reg: 0.321 loss_rpn_cls: 0.000 loss_rpn_loc: 0.010 time: 2.2509 data_time: 0.0087 lr: 0.000899 max_mem: 8426M [09/04 19:03:25 d2.utils.events]: eta: 0:03:02 iter: 919 total_loss: 0.361 loss_cls: 0.063 loss_box_reg: 0.287 loss_rpn_cls: 0.001 loss_rpn_loc: 0.009 time: 2.2504 data_time: 0.0085 lr: 0.000919 max_mem: 8426M [09/04 19:04:10 d2.utils.events]: eta: 0:02:17 iter: 939 total_loss: 0.367 loss_cls: 0.068 loss_box_reg: 0.278 loss_rpn_cls: 0.000 loss_rpn_loc: 0.010 time: 2.2507 data_time: 0.0095 lr: 0.000939 max_mem: 8426M [09/04 19:04:56 d2.utils.events]: eta: 0:01:32 iter: 959 total_loss: 0.385 loss_cls: 0.059 loss_box_reg: 0.298 loss_rpn_cls: 0.001 loss_rpn_loc: 0.009 time: 2.2510 data_time: 0.0091 lr: 0.000959 max_mem: 8426M [09/04 19:05:41 d2.utils.events]: eta: 0:00:47 iter: 979 total_loss: 0.349 loss_cls: 0.063 loss_box_reg: 0.283 loss_rpn_cls: 0.000 loss_rpn_loc: 0.009 time: 2.2511 data_time: 0.0090 lr: 0.000979 max_mem: 8426M WARNING [09/04 19:06:27 d2.data.datasets.coco]: Category ids in annotations are not in [1, #categories]! We'll apply a mapping for you. [09/04 19:06:27 d2.data.datasets.coco]: Loaded 10 images in COCO format from /content/valid/_annotations.coco.json [09/04 19:06:27 d2.data.common]: Serializing 10 elements to byte tensors and concatenating them all ... [09/04 19:06:27 d2.data.common]: Serialized dataset takes 0.00 MiB [09/04 19:06:27 d2.evaluation.evaluator]: Start inference on 10 images [09/04 19:06:29 d2.evaluation.evaluator]: Total inference time: 0:00:01.069654 (0.213931 s / img per device, on 1 devices) [09/04 19:06:29 d2.evaluation.evaluator]: Total inference pure compute time: 0:00:00 (0.198577 s / img per device, on 1 devices) [09/04 19:06:29 d2.evaluation.coco_evaluation]: Preparing results for COCO format ... [09/04 19:06:29 d2.evaluation.coco_evaluation]: Saving results to coco_eval/coco_instances_results.json [09/04 19:06:29 d2.evaluation.coco_evaluation]: Evaluating predictions ... Loading and preparing results... DONE (t=0.00s) creating index... index created! Running per image evaluation... Evaluate annotation type *bbox* DONE (t=0.01s). Accumulating evaluation results... DONE (t=0.01s). Average Precision (AP) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = 0.448 Average Precision (AP) @[ IoU=0.50 | area= all | maxDets=100 ] = 0.723 Average Precision (AP) @[ IoU=0.75 | area= all | maxDets=100 ] = 0.508 Average Precision (AP) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.448 Average Precision (AP) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = -1.000 Average Precision (AP) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = -1.000 Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets= 1 ] = 0.407 Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets= 10 ] = 0.493 Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = 0.493 Average Recall (AR) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.493 Average Recall (AR) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = -1.000 Average Recall (AR) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = -1.000 [09/04 19:06:29 d2.evaluation.coco_evaluation]: Evaluation results for bbox: | AP | AP50 | AP75 | APs | APm | APl | |:------:|:------:|:------:|:------:|:-----:|:-----:| | 44.818 | 72.343 | 50.825 | 44.818 | nan | nan | [09/04 19:06:29 d2.evaluation.coco_evaluation]: Note that some metrics cannot be computed. [09/04 19:06:29 d2.evaluation.coco_evaluation]: Per-category bbox AP: | category | AP | category | AP | |:-----------|:-----|:-----------|:-------| | Car | nan | Car | 44.818 | [09/04 19:06:29 d2.engine.defaults]: Evaluation results for my_dataset_val in csv format: [09/04 19:06:29 d2.evaluation.testing]: copypaste: Task: bbox [09/04 19:06:29 d2.evaluation.testing]: copypaste: AP,AP50,AP75,APs,APm,APl [09/04 19:06:29 d2.evaluation.testing]: copypaste: 44.8185,72.3432,50.8251,44.8185,nan,nan [09/04 19:06:29 d2.utils.events]: eta: 0:00:02 iter: 999 total_loss: 0.361 loss_cls: 0.054 loss_box_reg: 0.304 loss_rpn_cls: 0.001 loss_rpn_loc: 0.008 time: 2.2494 data_time: 0.0085 lr: 0.000999 max_mem: 8426M [09/04 19:06:29 d2.engine.hooks]: Overall training speed: 997 iterations in 0:37:24 (2.2516 s / it) [09/04 19:06:29 d2.engine.hooks]: Total training time: 0:37:33 (0:00:08 on hooks)
In [ ]:
# Look at training curves in tensorboard:
%load_ext tensorboard
%tensorboard --logdir output
The tensorboard extension is already loaded. To reload it, use: %reload_ext tensorboard
Evaluation¶
In [ ]:
#test evaluation
cfg.MODEL.WEIGHTS = os.path.join(cfg.OUTPUT_DIR, "model_final.pth")
cfg.MODEL.ROI_HEADS.SCORE_THRESH_TEST = 0.85 # set the test threshold
predictor = DefaultPredictor(cfg)
evaluator = COCOEvaluator("my_dataset_test", cfg, False, output_dir="./output/")
val_loader = build_detection_test_loader(cfg, "my_dataset_test")
inference_on_dataset(trainer.model, val_loader, evaluator)
WARNING [09/04 19:08:05 d2.data.datasets.coco]: Category ids in annotations are not in [1, #categories]! We'll apply a mapping for you. [09/04 19:08:05 d2.data.datasets.coco]: Loaded 40 images in COCO format from /content/test/_annotations.coco.json [09/04 19:08:05 d2.data.common]: Serializing 40 elements to byte tensors and concatenating them all ... [09/04 19:08:05 d2.data.common]: Serialized dataset takes 0.01 MiB [09/04 19:08:05 d2.evaluation.evaluator]: Start inference on 40 images [09/04 19:08:07 d2.evaluation.evaluator]: Inference done 11/40. 0.1975 s / img. ETA=0:00:05 [09/04 19:08:13 d2.evaluation.evaluator]: Inference done 36/40. 0.1998 s / img. ETA=0:00:00 [09/04 19:08:13 d2.evaluation.evaluator]: Total inference time: 0:00:07.126319 (0.203609 s / img per device, on 1 devices) [09/04 19:08:13 d2.evaluation.evaluator]: Total inference pure compute time: 0:00:06 (0.199904 s / img per device, on 1 devices) [09/04 19:08:13 d2.evaluation.coco_evaluation]: Preparing results for COCO format ... [09/04 19:08:13 d2.evaluation.coco_evaluation]: Saving results to ./output/coco_instances_results.json [09/04 19:08:13 d2.evaluation.coco_evaluation]: Evaluating predictions ... Loading and preparing results... DONE (t=0.00s) creating index... index created! Running per image evaluation... Evaluate annotation type *bbox* DONE (t=0.06s). Accumulating evaluation results... DONE (t=0.01s). Average Precision (AP) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = 0.326 Average Precision (AP) @[ IoU=0.50 | area= all | maxDets=100 ] = 0.728 Average Precision (AP) @[ IoU=0.75 | area= all | maxDets=100 ] = 0.204 Average Precision (AP) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.329 Average Precision (AP) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.000 Average Precision (AP) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = -1.000 Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets= 1 ] = 0.164 Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets= 10 ] = 0.453 Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = 0.468 Average Recall (AR) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.472 Average Recall (AR) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.000 Average Recall (AR) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = -1.000 [09/04 19:08:14 d2.evaluation.coco_evaluation]: Evaluation results for bbox: | AP | AP50 | AP75 | APs | APm | APl | |:------:|:------:|:------:|:------:|:-----:|:-----:| | 32.581 | 72.802 | 20.416 | 32.923 | 0.000 | nan | [09/04 19:08:14 d2.evaluation.coco_evaluation]: Note that some metrics cannot be computed. [09/04 19:08:14 d2.evaluation.coco_evaluation]: Per-category bbox AP: | category | AP | category | AP | |:-----------|:-----|:-----------|:-------| | Car | nan | Car | 32.581 |
Out[ ]:
OrderedDict([('bbox',
{'AP': 32.58125848458507,
'AP50': 72.80156797550605,
'AP75': 20.416164316726025,
'APs': 32.92277916692182,
'APm': 0.0,
'APl': nan,
'AP-Car': 32.58125848458507})])
Inference with the model saved weights¶
In [ ]:
%ls ./output/
coco_instances_results.json last_checkpoint events.out.tfevents.1662305966.2bcf4c9aba09.76.0 metrics.json events.out.tfevents.1662316132.2bcf4c9aba09.76.1 model_final.pth instances_predictions.pth
Test the model on my custom dataset¶
In [ ]:
cfg.MODEL.WEIGHTS = os.path.join(cfg.OUTPUT_DIR, "model_final.pth")
cfg.DATASETS.TEST = ("my_dataset_test", )
cfg.MODEL.ROI_HEADS.SCORE_THRESH_TEST = 0.7 # set the testing threshold for this model
predictor = DefaultPredictor(cfg)
test_metadata = MetadataCatalog.get("my_dataset_test")
The desired algorithm¶
Convert image to green¶
In [ ]:
def convert_image_to_green(image: np.ndarray) -> np.ndarray:
img_green = image.copy() # Create a copy of the image
# Cancel out the other colors (Blue and Red)
img_green[:, :, 0] = 0
img_green[:, :, 2] = 0
return img_green
paint bboxes with red¶
In [ ]:
def paint_bbox(image: np.ndarray, bbox: torch.tensor):
image_with_boxes = image.copy() # Create a copy of the image
total_area = 0
for box in bbox: # All predictions in the image
x0, y0, x1, y1 = box # top left & bottom right coords
x0 = int(x0)
y0 = int(y0)
x1 = int(x1)
y1 = int(y1)
w = x1 - x0
h = y1 - y0
total_area += w*h
points = np.array([[x0, y0], [x0, y1], [x1, y1], [x1, y0]])
image_with_boxes = cv2.fillPoly(image_with_boxes, pts=[points], color=(0,0,255)) # paint the box red
return image_with_boxes, total_area
The scan and paint algorithm¶
In [ ]:
def scan_and_paint(img: np.ndarray):
res = img.copy()
outputs = predictor(res) # predict bboxes for the image
res = convert_image_to_green(res) #convert to green
res, cars_area = paint_bbox(res, outputs["instances"].pred_boxes.tensor) # paint bboxes with red & get cars area in the image
img_area = img.shape[0] * img.shape[1]
print(f'The cars area is: {cars_area:.3f}')
print(f'The image area is: {img_area:.3f}')
print(f'The cars percentage in the image is: {(cars_area/img_area):.3f}%')
cv2_imshow(res)
In [ ]:
for imageName in glob.glob('/content/test/*jpg'):
im = cv2.imread(imageName)
scan_and_paint(im)
The cars area is: 927.000 The image area is: 173056.000 The cars percentage in the image is: 0.005%
The cars area is: 195.000 The image area is: 173056.000 The cars percentage in the image is: 0.001%
The cars area is: 0.000 The image area is: 173056.000 The cars percentage in the image is: 0.000%
The cars area is: 539.000 The image area is: 173056.000 The cars percentage in the image is: 0.003%
The cars area is: 571.000 The image area is: 173056.000 The cars percentage in the image is: 0.003%
The cars area is: 2464.000 The image area is: 173056.000 The cars percentage in the image is: 0.014%
The cars area is: 400.000 The image area is: 173056.000 The cars percentage in the image is: 0.002%
The cars area is: 765.000 The image area is: 173056.000 The cars percentage in the image is: 0.004%
The cars area is: 2221.000 The image area is: 173056.000 The cars percentage in the image is: 0.013%
The cars area is: 893.000 The image area is: 173056.000 The cars percentage in the image is: 0.005%
The cars area is: 679.000 The image area is: 173056.000 The cars percentage in the image is: 0.004%
The cars area is: 2765.000 The image area is: 173056.000 The cars percentage in the image is: 0.016%
The cars area is: 360.000 The image area is: 173056.000 The cars percentage in the image is: 0.002%
The cars area is: 2736.000 The image area is: 173056.000 The cars percentage in the image is: 0.016%
The cars area is: 210.000 The image area is: 173056.000 The cars percentage in the image is: 0.001%
The cars area is: 345.000 The image area is: 173056.000 The cars percentage in the image is: 0.002%
The cars area is: 196.000 The image area is: 173056.000 The cars percentage in the image is: 0.001%
The cars area is: 253.000 The image area is: 173056.000 The cars percentage in the image is: 0.001%
The cars area is: 2379.000 The image area is: 173056.000 The cars percentage in the image is: 0.014%
The cars area is: 970.000 The image area is: 173056.000 The cars percentage in the image is: 0.006%
The cars area is: 583.000 The image area is: 173056.000 The cars percentage in the image is: 0.003%
The cars area is: 1746.000 The image area is: 173056.000 The cars percentage in the image is: 0.010%
The cars area is: 2590.000 The image area is: 173056.000 The cars percentage in the image is: 0.015%
The cars area is: 569.000 The image area is: 173056.000 The cars percentage in the image is: 0.003%
The cars area is: 2443.000 The image area is: 173056.000 The cars percentage in the image is: 0.014%
The cars area is: 168.000 The image area is: 173056.000 The cars percentage in the image is: 0.001%
The cars area is: 1236.000 The image area is: 173056.000 The cars percentage in the image is: 0.007%
The cars area is: 492.000 The image area is: 173056.000 The cars percentage in the image is: 0.003%
The cars area is: 2982.000 The image area is: 173056.000 The cars percentage in the image is: 0.017%
The cars area is: 1393.000 The image area is: 173056.000 The cars percentage in the image is: 0.008%
The cars area is: 506.000 The image area is: 173056.000 The cars percentage in the image is: 0.003%
The cars area is: 934.000 The image area is: 173056.000 The cars percentage in the image is: 0.005%
The cars area is: 1532.000 The image area is: 173056.000 The cars percentage in the image is: 0.009%
The cars area is: 2755.000 The image area is: 173056.000 The cars percentage in the image is: 0.016%
The cars area is: 543.000 The image area is: 173056.000 The cars percentage in the image is: 0.003%
The cars area is: 984.000 The image area is: 173056.000 The cars percentage in the image is: 0.006%
The cars area is: 247.000 The image area is: 173056.000 The cars percentage in the image is: 0.001%
The cars area is: 2348.000 The image area is: 173056.000 The cars percentage in the image is: 0.014%
The cars area is: 480.000 The image area is: 173056.000 The cars percentage in the image is: 0.003%
The cars area is: 844.000 The image area is: 173056.000 The cars percentage in the image is: 0.005%
- An option to see the images & predictions in RGB while using detectron2 instances drawing
In [ ]:
for imageName in glob.glob('/content/test/*jpg'):
im = cv2.imread(imageName)
outputs = predictor(im)
v = Visualizer(im[:, :, ::-1],
metadata=test_metadata,
scale=0.8
)
out = v.draw_instance_predictions(outputs["instances"].to("cpu"))
cv2_imshow(out.get_image()[:, :, ::-1])
